CN102340811B - Method for carrying out fault diagnosis on wireless sensor networks - Google Patents

Method for carrying out fault diagnosis on wireless sensor networks Download PDF

Info

Publication number
CN102340811B
CN102340811B CN201110342475.5A CN201110342475A CN102340811B CN 102340811 B CN102340811 B CN 102340811B CN 201110342475 A CN201110342475 A CN 201110342475A CN 102340811 B CN102340811 B CN 102340811B
Authority
CN
China
Prior art keywords
node
fault
amp
neighbor
wireless sensor
Prior art date
Application number
CN201110342475.5A
Other languages
Chinese (zh)
Other versions
CN102340811A (en
Inventor
李道亮
李秋成
李振波
马道坤
丁启胜
王振智
魏晓华
Original Assignee
中国农业大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国农业大学 filed Critical 中国农业大学
Priority to CN201110342475.5A priority Critical patent/CN102340811B/en
Publication of CN102340811A publication Critical patent/CN102340811A/en
Application granted granted Critical
Publication of CN102340811B publication Critical patent/CN102340811B/en

Links

Abstract

The invention relates to the field of wireless sensor networks and fault diagnosis, and discloses a method for carrying out fault diagnosis on wireless sensor networks. The method comprises the following steps: S1, carrying out online real-time acquisition on data measured by sensor nodes in a wireless sensor network; S2, detecting the fault of a single sensor node by using the space and time characteristics of the data; S3, detecting the fault of the whole wireless sensor network by using the correlation between two adjacent nodes; S4, carrying out attribute reduction on parameters obtained in the steps S2 and S3; and S5, obtaining a fault diagnosis result of the wireless sensor network according to the attribute reduction result obtained in the step S4. In the invention, the fault of a single sensor node is detected by using the space and time characteristics of data measured by sensor nodes in a wireless sensor network, and the rapid and correct fault detection on all nodes in the network by using the correlation between the nodes is realized.

Description

Wireless sensor network fault diagnosis method

Technical field

The present invention relates to wireless sensor network and fault diagnosis field, particularly a kind of wireless sensor network fault diagnosis method that is applicable to indoor aquaculture.

Background technology

From the research to indoor aquaculture water quality supervisory control system both at home and abroad, relatively ripe to the research of the indoor aquaculture water quality supervisory control system based on cable technology at present, but while applying this system, indoorly a large amount of signal transmssion lines and power line must be covered with, cause facility build difficulty, I&M is complicated, workload is large, cost is high, reliability is low, has limited the application of the mobile operating equipment such as agricultural robot simultaneously.Existing automation, intellectuality, " unmanned " degree that is unfavorable for further improving the indoor aquaculture of modernization based on the indoor aquaculture water quality supervisory control system of cable technology.It is the important research field of indoor aquaculture water quality supervisory control system in recent years that wireless sensor network is applied to indoor aquatic products TT&C system.The research of wireless sensor network (WSN) failure diagnosis to indoor aquaculture water quality supervisory control system is still rare.

Because the WSN of indoor aquaculture water quality supervisory control system works in hot and humid water environment, and sensor node in WSN is general battery-powered and by wireless communication mode transceiving data, have the shortcomings such as finite energy and antijamming capability be weak, the possibility that therefore sensor node breaks down is higher.Once the WSN of indoor aquaculture pond water quality monitoring system produces fault detection failure source processing in time again, just likely cause network paralysis, reduce system reliability, even cause TT&C system out of control, jeopardize the production safety in whole indoor culture pond.Therefore how in real time the running status of the WSN to indoor wireless water quality monitoring system is carried out failure diagnosis, finds in time fault and locate the source of trouble just to have become one of WSN priority fields of study of indoor aquaculture water quality supervisory control system.

Summary of the invention

(1) technical problem that will solve

Technical problem to be solved by this invention is: how to realize all quick, correct fault detects of node in wireless sensor network.

(2) technical scheme

For solving the problems of the technologies described above, the invention provides a kind of wireless sensor network fault diagnosis method, comprise the following steps:

The measured data of sensor node in S1, online real time collecting wireless sensor network;

S2, utilize the space-time characterisation of described data to detect the fault of single-sensor node;

S3, utilize the fault of the whole wireless sensor network of correlation detection before adjacent node;

S4, to step S2 and these two Procedure Acquisitions of step S3 to parameter carry out attribute reduction;

S5, draw the fault diagnosis result of wireless sensor network according to the attribute reduction result of step S4.

Preferably, in step S2 by settling time sequence fault diagnosis model and spatial sequence fault diagnosis model detect the fault of single-sensor node.

Preferably, in step S3, be adjacent the number that difference between the measured data of sensor node exceedes threshold value and judge whether present node breaks down by detecting current sensor node.

Preferably, in step S4, the parameter of utilizing coarse central algorithm to get step S2 and step S3 is carried out attribute reduction.

Preferably, step S2 specifically comprises:

201, the data in each sampling period of sensor node are respectively organized in preliminary treatment, obtain diagnostic sample, and the diagnostic sample of establishing A group is A ' i, h+t, make h=0, wherein, i represents the sequence number of sensor node in place group, and the aggregation node of establishing A group is also A, and in this group, the identifier of all sensors node is pressed the sequence of positions called after A at place separately 1, A 2..., A p, wherein p is the sum of this group inner sensor node, t is the collection period of t transducer;

202, select time Series Modeling data j, to time series A ' i, h+1, A ' i, h+2..., A ' i, h+jcarry out modeling, obtain y (h+j) and Y r, δ is the standard deviation of the model residual error of time series diagnostic model; ε is model residual error; Y is the trend factor; Y rbe factor variations threshold value, make variable q=1;

203, to diagnostic sample A ' i, h+j+qcarry out time series analysis, obtain y (h+j+q) and B i, h+j+q, sensor node A when B is illustrated in i+q sampling period itime series operating state, B i, h+j+q=Y (h+j)-Y r;

204, work as B i, h+j+q≤ 0 o'clock, q is added to 1, again perform step 203; Work as B i, h+j+qwhen > 0, execution step 205;

205,, according to node location, determine the value of spatial sequence mark C, then h+j+q cycle A organized to the diagnostic sample A ' of all sensors node 1, h+j+q, A ' 2, h+j+q..., A ' p, h+j+qdetermine spatial sequence k to be analyzed 1, k 2..., k p;

206, to spatial sequence k 1, k 2..., k pcarry out modeling, obtain z (1) and Z r;

Spatial sequence fault diagnosis model is:

In formula ---auto-regressive parameter, s=1,2, ..., f;

F---Autoregressive;

Model output residual sequence ε ' rvariance be:

The trend factor is:

Z ( r ) = δ ϵ ′ 2 ( r ) - δ ϵ ′ 2 ( r - 1 )

Making spatial sequence modeling data amount is l, factor variations threshold value Z rfor

Wherein, for | Z (f+2) |, | Z (f+3) | ..., | Z (1) | average,

δ zfor | Z (f+2) |, | Z (f+3) | ..., | Z (1) | standard deviation;

W isensor node A while representing m sampling period ioperating state;

W i=Z(p-i+1)-Z R

Work as W i≤ 0 represents A iworking properly, work as W i> 0 represents A ibreak down;

207, in the time of C=0 to k icarry out spatial sequence analysis, obtain z (i) and W i;

In the time of C=1 to k p-i+1carry out spatial sequence analysis, obtain z (p-i+1) and W i;

208, work as W iwhen > 0, send warning, and point out the position of malfunctioning node; Work as W i≤ 0 o'clock is h+j+q by h assignment, returns to execution step 202.

Preferably, step S3 specifically comprises:

By node A ithe set of all adjacent nodes be designated as Neighbor (A i); A ineighbor node sum, be designated as Num (Neighbor (A i)), the difference of the data that moment t measures separately be no more than threshold value θ 1; And at another moment t+1, the measured data of two adjacent nodes poor with difference be no more than threshold value θ 2;

301, consider Neighbor (A i) in arbitrary node A i, put C ij=0, calculate c ijrepresent test result, if node A iwith A jin have at least one to break down, make test result C ij=1, otherwise make C ij=0;

If put C ij=1, go to Neighbor (A i) in next node;

If calculate if put C ij=1, go to Neighbor (A i) in next node;

Repeat above-mentioned steps, until obtain A iwith Neighbor (A i) in each internodal test result;

If 302 &Sigma; A j &Element; Neighbor ( A i ) C ij < Num ( Neighbor ( A i ) ) / 2 , Make A itentative diagnosis state T ifor the normal LG of possibility, otherwise T ifor possible breakdown LT;

303, order for S ineighbor node in the tentative diagnosis state node number that is LG, if

&Sigma; A j &Element; Neighbor ( A i ) and T i = LG C ij < Num ( Neighbor ( S i ) T j = LG ) / 2 , Make A istate be normal GD, otherwise be fault FT;

If 304 node A ineighbor node in tentative diagnosis state be LG nodes is 0, further, if A itentative diagnosis state T ifor LG, make A istate be normal GD, otherwise be fault FT;

305, check whether completed the condition diagnosing to all nodes of network, if complete, exit, otherwise repeating step 301~304.

Preferably, described coarse central algorithm is as follows:

401, establishing feedforward layer connection weights in Hamming neural net is wherein i=1,2 ..., p; J '=1,2 ..., n, represent i master sample pattern vector c ithe individual element value of j '; In feedforward layer, the threshold values of each processing unit is made as θ i=-n/2 (i=1,2 ... p), the activation primitive of feedforward layer is made as f 1(x)=x/n, and establish t=0, j ' represents the individual element of j ' in each sample, its size is no more than the number n of the fault attribute that node likely exists;

402, after feedforward layer r i ( t ) = ( &Sigma; i = 1 n w j &prime; i x j &prime; - &theta; i ) / n , i = 1,2 , . . . p ;

403, establish in Hamming neural net the threshold values of processing unit in competition layer and be 0, activation primitive is:

f ( k ) = o , k < 0 k , k &GreaterEqual; 0

404, appoint and get an e value that meets o < e < 1/ (p-1);

405, calculate r i ( t + 1 ) = f ( r i ( t ) - e &Sigma; m &NotEqual; i r m ( t ) ) , i = 1,2 , . . . p , The span of m is 1,2 ... in p, do not comprise the value of i;

406, calculate &delta; = &Sigma; i = 1 p ( r i ( t ) - r i ( t + 1 ) ) ;

If 407 δ ≠ 0, add 1 by t, and forward step 405 to;

408, output r j '(t+1) be on the occasion of item, be the classification that x is corresponding, the x kind of fault existing that expresses possibility.

(3) beneficial effect

The present invention utilizes the space-time characterisation of the measured data of the sensor node in wireless sensor network to detect the fault of individual node, and utilize correlation between node realize in network whole nodes fast, correctly fault detect.Also and according to the applied environment of wireless sensor network and fault signature obtain diagnosing decision table, utilize the conclusion old attribute reduction algorithms in improved rough set to carry out attribute reduction to diagnosis decision table, set up the method for a set of failure modes with Hamming network.

Brief description of the drawings

Fig. 1 is the method flow diagram of the embodiment of the present invention;

Fig. 2 is the indoor aquaculture WSN topological structure schematic diagram that the embodiment of the present invention provides.

Embodiment

Under regard to a kind of wireless sensor network fault diagnosis method proposed by the invention, in conjunction with the accompanying drawings and embodiments describe in detail.

As shown in Figure 1, the invention provides a kind of wireless sensor network fault diagnosis method that is applicable to indoor aquaculture, comprise the following steps:

The measured data of sensor node in S1, online real time collecting wireless sensor network (in the present invention also referred to as " transducer ").Transducer is divided into a variety of, refers in embodiments of the present invention the water quality sensor such as dissolved oxygen sensor, temperature sensor, and corresponding data are exactly the data such as dissolved oxygen, water temperature in the water adopted of transducer.

S2, utilize the space-time characterisation of described data to detect the fault of single-sensor node.In this step by settling time sequence fault diagnosis model and spatial sequence fault diagnosis model detect.

The indoor aquaculture WSN individual node fault diagnosis algorithm based on the two sequences of space-time of this step specifically describes as follows:

201) data in each sampling period of sensor node are respectively organized in preliminary treatment, obtain diagnostic sample.If the diagnostic sample of A group is A ' i, h+t, make h=0, wherein, i represents the sequence number of sensor node in place group.If the aggregation node of A group is also A.In this group, the identifier of all sensors node is pressed the sequence of positions called after A at place separately 1, A 2..., A p, wherein p is the sum of this group inner sensor node.The implication of t is the collection period of t transducer.

202) according to indoor aquaculture pond particular case, (the various environmental informations that indoor aquaculture WSN collects show as and change slowly, the time series of information structure and spatial sequence stationarity are comparatively obvious), select time Series Modeling data j, to time series A ' i, h+1, A ' i, h+2..., A ' i, h+jcarry out modeling, obtain y (h+j) and Y r.The implication of δ is the standard deviation of the model residual error of time series diagnostic model; The implication of ε is model residual error; The implication of Y is the trend factor; Y rimplication be factor variations threshold value, make variable q=1.

203) to diagnostic sample A ' i, h+j+qcarry out time series analysis, obtain y (h+j+q) and B i, h+j+q.Sensor node A when B is illustrated in i+q sampling period itime series operating state, B i, h+j+q=Y (h+j)-Y r.

204) work as B i, h+j+q≤ 0 o'clock, q is added to 1, again perform step 203; Work as B i, h+j+qwhen > 0, carry out next step.

205), according to node location, determine the value of spatial sequence mark C, then h+j+q cycle A organized to the diagnostic sample A ' of all sensors node 1, h+j+q, A ' 2, h+j+q..., A ' p, h+j+qdetermine spatial sequence k to be analyzed 1, k 2..., k p.

206) to spatial sequence k 1, k 2..., k pcarry out modeling, obtain z (1) and Z r.

Spatial sequence fault diagnosis model is:

In formula ---auto-regressive parameter, s=1,2, ..., f;

F---Autoregressive;

The implication of r is the modeling data amount of spatial sequence fault diagnosis model, and the span of r is 0 < r≤p, and p is the total data volume of spatial sequence.

Model output residual sequence ε ' rvariance be:

The trend factor is:

Z ( r ) = &delta; &epsiv; &prime; 2 ( r ) - &delta; &epsiv; &prime; 2 ( r - 1 )

Making spatial sequence modeling data amount is l, factor variations threshold value Z rfor

Wherein, for | Z (f+2) |, | Z (f+3) | ..., | Z (1) | average,

δ zfor | Z (f+2) |, | Z (f+3) | ..., | Z (1) | standard deviation.

W isensor node A while representing m sampling period ioperating state

W i=Z(p-i+1)-Z R

Work as W i≤ 0 represents A iworking properly, work as W i> 0 represents A ibreak down.

207) in the time of C=0 to k icarry out spatial sequence analysis, obtain z (i) and W i;

In the time of C=1 to k p-i+1carry out spatial sequence analysis, obtain z (p-i+1) and W i.

208) work as W iwhen > 0, send warning, and point out the position of malfunctioning node; Work as W i≤ 0 o'clock is h+j+q by h assignment, returns to execution step 202.

S3, utilize the fault of the whole wireless sensor network of correlation detection before adjacent node; The number that exceedes threshold value with the difference of data that its adjacent node detects by detection present node judges whether present node breaks down, and described threshold value is analyze by experiment and add up the empirical value obtaining.If in the communication range in a jumping between two sensor nodes, think that they are adjacent nodes, are neighbor node each other.

Certain node A ithe set of all adjacent nodes, be designated as Neighbor (A i); A ineighbor node sum, be designated as Num (Neighbor (A i)).Because two nodes are adjacent, distance is very near, certain moment t measures the data that (or claim perception) arrive separately should be very approaching, its difference be no more than a certain threshold value θ 1; And at another moment t+1, the data of two adjacent node institutes perception poor with difference can not differ too large, be no more than a certain threshold value θ yet 2.

301) consider Neighbor (A i) in arbitrary node A i, put C ij=0 calculates c ijrepresent test result, if node A iwith A jin have at least one to break down, make test result C ij=1, otherwise make C ij=0.

If put C ij=1, go to Neighbor (A i) in next node.

If calculate if put C ij=1, go to Neighbor (A i) in next node.

Repeat above-mentioned steps, until obtain A iwith Neighbor (A i) in each internodal test result.

302) if &Sigma; A j &Element; Neighbor ( A i ) C ij < Num ( Neighbor ( A i ) ) / 2 , Make A itentative diagnosis state T ifor possibility normal (LG), otherwise T ifor possible breakdown (LT).

303) order for S ineighbor node in the tentative diagnosis state node number that is LG, if

&Sigma; A j &Element; Neighbor ( A i ) and T i = LG C ij < Num ( Neighbor ( S i ) T j = LG ) / 2 , Make A istate be normal (GD), otherwise be fault (FT).

304) if node A ineighbor node in tentative diagnosis state be LG nodes is 0, further, if A itentative diagnosis state T ifor LG, make A istate be normal (GD), otherwise be fault (FT).

305) check whether completed the condition diagnosing to all nodes of network, if complete, exit, otherwise repeating step 301,302,303,304.

S4, utilize coarse central algorithm to step S2 and these two Procedure Acquisitions of step S3 to parameter carry out attribute reduction.Attribute reduction refers under guarantee information genealogical classification or the constant condition of decision-making capability, deletes the redundant attributes in conditional attribute, thereby reduces data mining data volume to be processed, improves the terseness of data mining results.Coarse central algorithm is as follows:

401) establishing feedforward layer connection weights in Hamming neural net is wherein i=1,2 ..., p; J '=1,2 ..., n, represent i master sample pattern vector c ithe individual element value of j '; In feedforward layer, the threshold values of each processing unit is made as θ i=-n/2 (i=1,2 ... p), the activation primitive of feedforward layer is made as f 1(x)=x/n.And establish t=0.J ' represents the individual element of j ' in each sample, and its size can not exceed the number n of the fault attribute that node likely exists.

In sensor network, except transducer, also have wireless collection device, the transmission equipments such as GPRS gateway, if these device fails also can have influence on the judgement to sensor states.In preceding step S2, S3 in the processing procedure of sensing data, also can obtain the state information of wireless collection device and GPRS gateway, state information is divided into again communications status and energy state, like this attribute is just many, and some is related before these attributes, some does not have association.In order to obtain a more definite fault point, method of the present invention is not have related attribute to leaving out by Hamming neural net handle, thereby obtains more definite fault point.

402) after feedforward layer r i ( t ) = ( &Sigma; i = 1 n w j &prime; i x j &prime; - &theta; i ) / n , i = 1,2 , . . . p , The implication of this formula and wherein symbol is prior art, can be referring to Hamming neural network theory knowledge.

403) establish in Hamming neural net the threshold values of processing unit in competition layer and be 0, activation primitive is:

f ( k ) = o , k < 0 k , k &GreaterEqual; 0

404) appoint and get an e value that meets o < e < 1/ (p-1).

405) calculate r i ( t + 1 ) = f ( r i ( t ) - e &Sigma; m &NotEqual; i r m ( t ) ) , i = 1,2 , . . . p . The span of m is 1,2 ... in p, do not comprise the value of i;

406) calculate &delta; = &Sigma; i = 1 p ( r i ( t ) - r i ( t + 1 ) ) .

407) if δ ≠ 0 adds 1 by t, and forward step 405 to.

408) output r j '(t+1) be on the occasion of item, be the classification that x is corresponding.X represents the one in fault that transducer, wireless collection device, all these equipment of GPRS gateway may exist.

In Hamming neural net, feedforward layer is identical with the number of classification samples with the neuron number of competition layer, and neuronic position is corresponding with concrete classification.The essence of this algorithm be parallel computation n deduct each Hamming distance from value.Hamming neural net can be adjusted to the fault mode classification device with minimal error, and it can provide a fault mode and come the input fault pattern of matched position, there will not be unmatched result.

The attribute reduction result of S5, step S4 is whether the equipment of judging in wireless sensor network has fault, if out of order word, specifically sensor fault, or the fault of wireless collection device, or the fault of which equipment in these equipment of GPRS gateway.Draw the fault diagnosis result of wireless sensor network according to the attribute reduction result of step S4.

Above execution mode is only for illustrating the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (3)

1. a wireless sensor network fault diagnosis method, is characterized in that, comprises the following steps:
The measured data of sensor node in S1, online real time collecting wireless sensor network;
S2, utilize the space-time characterisation of described data to detect the fault of single-sensor node;
S3, utilize the fault of the whole wireless sensor network of correlation detection between adjacent node;
S4, to step S2 and these two Procedure Acquisitions of step S3 to parameter carry out attribute reduction;
S5, draw the fault diagnosis result of wireless sensor network according to the attribute reduction result of step S4;
Wherein, in step S2 by settling time sequence fault diagnosis model and spatial sequence fault diagnosis model detect the fault of single-sensor node;
Wherein, in step S3, be adjacent the number that difference between the measured data of sensor node exceedes threshold value and judge whether present node breaks down by detecting current sensor node;
Step S2 specifically comprises:
201, the data in each sampling period of sensor node are respectively organized in preliminary treatment, obtain diagnostic sample, and the diagnostic sample of establishing A group is A' i, h+t, make h=0, wherein, i represents the sequence number of sensor node in place group, and the aggregation node of establishing A group is also A, and in this group, the identifier of all sensors node is pressed the sequence of positions called after A at place separately 1, A 2..., A p, wherein p is the sum of this group inner sensor node, t is the collection period of t transducer;
202, select time Series Modeling data j, to time series A' i, h+1, A' i, h+2..., A' i, h+jcarry out modeling, obtain y (h+j) and Y r, δ is the standard deviation of the model residual error of time series diagnostic model; ε is model residual error; Y is the trend factor; Y rbe factor variations threshold value, make variable q=1;
203, to diagnostic sample A' i, h+j+qcarry out time series analysis, obtain y (h+j+q) and B i, h+j+q, sensor node A when B is illustrated in i+q sampling period itime series operating state, B i, h+j+q=Y (h+j)-Y r;
204, work as B i, h+j+q≤ 0 o'clock, q is added to 1, again perform step 203; Work as B i, h+j+qwhen >0, execution step 205;
205,, according to node location, determine the value of spatial sequence mark C, then h+j+q cycle A organized to the diagnostic sample A' of all sensors node 1, h+j+q, A' 2, h+j+q..., A' p, h+j+qdetermine spatial sequence k to be analyzed 1, k 2..., k p;
206, to spatial sequence k 1, k 2..., k pcarry out modeling, obtain z (1) and Z r;
Spatial sequence fault diagnosis model is:
In formula ---auto-regressive parameter, s=1,2 ..., f;
F---Autoregressive;
Model output residual sequence variance be:
The trend factor is:
Making spatial sequence modeling data amount is l, factor variations threshold value Z rfor
Wherein, for | Z (f+2) |, | Z (f+3) | ..., | Z (l) | average,
δ zfor | Z (f+2) |, | Z (f+3) | ..., | Z (l) | standard deviation;
W isensor node A while representing m sampling period ioperating state;
W i=Z(p-i+1)-Z R
Work as W i≤ 0 represents A iworking properly, work as W i>0 represents A ibreak down;
207, in the time of C=0 to k icarry out spatial sequence analysis, obtain z (i) and W i; In the time of C=1 to k p-i+1carry out spatial sequence analysis, obtain z (p-i+1) and W i;
208, work as W iwhen >0, send warning, and point out the position of malfunctioning node; Work as W i≤ 0 o'clock is h+j+q by h assignment, returns to execution step 202;
Step S3 specifically comprises:
By node A ithe set of all adjacent nodes be designated as Neighbor (A i); A ineighbor node sum, be designated as Num (Neighbor (A i)), the difference of the data that moment t measures separately be no more than threshold value θ 1; And at another moment t+1, the measured data of two adjacent nodes poor with difference be no more than threshold value θ 2;
301, consider Neighbor (A i) in arbitrary node A i, put C ij=0, calculate c ijrepresent test result, if node A iwith A jin have at least one to break down, make test result C ij=1, otherwise make C ij=0;
If put C ij=1, go to Neighbor (A i) in next node;
If calculate if put C ij=1, go to Neighbor (A i) in next node;
Repeat above-mentioned steps, until obtain A iwith Neighbor (A i) in each internodal test result;
If 302 &Sigma; C ij A j &Element; Neighbor ( A i ) < Num ( Neighbor ( A i ) ) / 2 , Make A itentative diagnosis state T ifor the normal LG of possibility, otherwise T ifor possible breakdown LT;
303, order for S ineighbor node in the tentative diagnosis state node number that is LG, if
&Sigma; C ij A j &Element; Neighbor ( A i ) and T i = LG < Num ( Neighbor ( S i ) T j = TG ) / 2 , Make A istate be normal GD, otherwise be fault FT;
If 304 node A ineighbor node in tentative diagnosis state be LG nodes is 0, further, if A itentative diagnosis state T ifor LG, make A istate be normal GD, otherwise be fault FT;
305, check whether completed the condition diagnosing to all nodes of network, if complete, exit, otherwise repeating step 301~304.
2. the method for claim 1, is characterized in that, in step S4, the parameter of utilizing coarse central algorithm to get step S2 and step S3 is carried out attribute reduction.
3. method as claimed in claim 2, is characterized in that, described coarse central algorithm is as follows:
401, establishing feedforward layer connection weights in Hamming neural net is wherein i=1,2 ..., p; J '=1,2 ..., n, represent i master sample pattern vector c ithe individual element value of j '; In feedforward layer, the threshold values of each processing unit is made as θ i=-n/2, i=1,2 ... p, the activation primitive of feedforward layer is made as f 1(x)=x/n, and establish t=0, j ' represents the individual element of j ' in each sample, its size is no more than the number n of the fault attribute that node likely exists;
402, after feedforward layer r i ( t ) = ( &Sigma; i = 1 n w j &prime; i x j &prime; - &theta; i ) / n , i = 1,2 , . . . p ;
403, establish in Hamming neural net the threshold values of processing unit in competition layer and be 0, activation primitive is:
404, appoint and get an e value that meets o<e<1/ (p-1);
405, calculate r i ( t + 1 ) = f ( r i ( t ) - e &Sigma; m &NotEqual; i r m ( t ) ) , i = 1,2 , . . . p , The span of m is 1,2 ... in p, do not comprise the value of i;
406, calculate &delta; = &Sigma; i = 1 p ( r i ( t ) - r i ( t + 1 ) ) ;
If 407 δ ≠ 0, add 1 by t, and forward step 405 to;
408, output r j'(t+1) be on the occasion of item, be the classification that x is corresponding, the x kind of fault existing that expresses possibility.
CN201110342475.5A 2011-11-02 2011-11-02 Method for carrying out fault diagnosis on wireless sensor networks CN102340811B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201110342475.5A CN102340811B (en) 2011-11-02 2011-11-02 Method for carrying out fault diagnosis on wireless sensor networks

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201110342475.5A CN102340811B (en) 2011-11-02 2011-11-02 Method for carrying out fault diagnosis on wireless sensor networks

Publications (2)

Publication Number Publication Date
CN102340811A CN102340811A (en) 2012-02-01
CN102340811B true CN102340811B (en) 2014-11-26

Family

ID=45516257

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201110342475.5A CN102340811B (en) 2011-11-02 2011-11-02 Method for carrying out fault diagnosis on wireless sensor networks

Country Status (1)

Country Link
CN (1) CN102340811B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107276531A (en) * 2017-06-08 2017-10-20 南昌航空大学 A kind of online failure sorted diagnostic system of photovoltaic module and method

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102680017A (en) * 2012-05-31 2012-09-19 潍柴动力股份有限公司 Fault diagnostic method and diagnostic device of sensor
CN102781027B (en) * 2012-07-18 2014-12-31 上海大学 Intelligent processing method for abnormal gateway in network activity period of wireless sensor
CN102970698A (en) * 2012-11-02 2013-03-13 北京邮电大学 Fault detection method of wireless sensor network
CN106325144A (en) * 2015-07-06 2017-01-11 湖州庆渔堂农业科技有限公司 Monitoring data calibration method and monitoring data calibration system of aquaculture monitoring system
CN105223945B (en) * 2015-09-22 2018-03-06 中国农业大学 Aquaculture Internet of Things equips the fault diagnosis method and system of cluster
CN105915388B (en) * 2016-05-31 2019-06-04 广东电网有限责任公司电力调度控制中心 A kind of Transducer-fault Detecting Method and system based on distributed network
CN105939524B (en) * 2016-06-21 2019-08-16 南京大学 A kind of wireless sensor network node event real-time predicting method
CN106255042A (en) * 2016-08-31 2016-12-21 烟台南山学院 A kind of chemical laboratory wireless sensor network node and method for diagnosing faults
CN107862375A (en) * 2017-10-30 2018-03-30 北京计算机技术及应用研究所 A kind of two stage equipment fault diagnosis method
CN108377261A (en) * 2018-01-24 2018-08-07 江苏新中天塑业有限公司 A kind of Diagnostic system of motor fault and method based on industrial wireless sensing network
CN108882192B (en) * 2018-07-19 2020-03-20 重庆邮电大学 Indoor area positioning method based on neighborhood rough set reduction

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101516099A (en) * 2009-04-07 2009-08-26 华中科技大学 Test method for sensor network anomaly

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101516099A (en) * 2009-04-07 2009-08-26 华中科技大学 Test method for sensor network anomaly

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
季赛,等.基于时空特性的无线传感器网络节点故障诊断方法.《传感器与微系统》.2009,第28卷(第10期),117-120. *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107276531A (en) * 2017-06-08 2017-10-20 南昌航空大学 A kind of online failure sorted diagnostic system of photovoltaic module and method
CN107276531B (en) * 2017-06-08 2019-08-27 南昌航空大学 A kind of online failure sorted diagnostic system of photovoltaic module and method

Also Published As

Publication number Publication date
CN102340811A (en) 2012-02-01

Similar Documents

Publication Publication Date Title
Lennox et al. Process monitoring of an industrial fed‐batch fermentation
WO2015158198A1 (en) Fault recognition method and system based on neural network self-learning
CN103914064B (en) Based on the commercial run method for diagnosing faults that multi-categorizer and D-S evidence merge
CN102289590B (en) Method for estimating operating state of SF6 high-voltage circuit breaker and intelligent system
CN203363353U (en) Process control valve assembly
CN102496069B (en) Cable multimode safe operation evaluation method based on fuzzy analytic hierarchy process (FAHP)
CN103336906B (en) The sampling Gaussian process regression model that in the image data stream of environmental sensor, continuous abnormal detects
CN103336507B (en) Automatically the statistical modeling divided based on the multi-modal collaborative period and on-line monitoring method
EP3017403A2 (en) System and method for abnormality detection
CN102175282B (en) Method for diagnosing fault of centrifugal air compressor based on information fusion
CN102520697B (en) Onsite information preprocessing method of remote cooperative diagnosis
CN103810328B (en) Transformer maintenance decision method based on hybrid model
CN102789545B (en) Based on the Forecasting Methodology of the turbine engine residual life of degradation model coupling
CN103302777B (en) Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine
Smarsly et al. Decentralized fault detection and isolation in wireless structural health monitoring systems using analytical redundancy
CN103268519B (en) Based on the power-system short-term load forecasting method and the device that improve Lyapunov index
CN101533058B (en) Power abnormal failure data analyzing device and diagnosing method
CN105511445B (en) Multi-modal procedure failure testing method based on local neighbor normalized matrix
CN105042339B (en) One kind is based on nondimensional leakage of finished oil pipeline amount estimating system and method
CN102122133B (en) Self-adaption wavelet neural network abnormity detection and fault diagnosis classification system and method
CN101382801B (en) Optimum soft measuring instrument based on EGA-optimized polymerization of propylene production process and method
CN101299004B (en) Vibrating failure diagnosis method based on determined learning theory
CN105046402B (en) A kind of state evaluating method applied to secondary equipment of intelligent converting station
CN101799367B (en) Electromechanical device neural network failure trend prediction method
CN102265227B (en) Method and apparatus for creating state estimation models in machine condition monitoring

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant